- Split View
-
Views
-
Cite
Cite
Yuhang Dai, Fanxi Gong, Xueqin Yang, Xiuzhi Chen, Yongxian Su, Liyang Liu, Jianping Wu, Xiaodong Liu, Qingling Sun, Litterfall seasonality and adaptive strategies of tropical and subtropical evergreen forests in China, Journal of Plant Ecology, Volume 15, Issue 2, April 2022, Pages 320–334, https://doi.org/10.1093/jpe/rtab102
- Share Icon Share
Abstract
Tropical and subtropical evergreen broad-leaved forests (EBFs) and needle-leaved forests (ENFs) in China exhibit complex leaf shedding strategies in responses to soil water availability, vapor pressure deficits (VPDs) and sunlight availability. However, the seasonal variations and triggers of litterfall differ significantly in tropical/subtropical forests, and there are still many uncertainties. Herein, we aim to explore the distinct climatic factors of seasonal litterfall in a climate–phenology correlation framework. We collected seasonal litterfall data from 85 sites across tropical/subtropical China and used linear correlation coefficients between sunlight and rainfall to partition synchronous/asynchronous climates. Additional phase analysis and structural equation model analysis were conducted to model the climatic triggers of tropical phenology. Results indicated two types of tropical litterfall phenology under two types of climates. In synchronous climates, where seasonal sunlight and rainfall are positively correlated, the litterfall peak of the unimodal phenology and the first litterfall peak of the bimodal phenology both happen at the end of dry season. The second litterfall peak of the bimodal phenology occurs at the end of rainy season due to water stress. In asynchronous climates, where seasonal sunlight and rainfall are negatively correlated, VPD shows consistent seasonal variations with incoming sunlight. The leaf senescence is accelerated at the end of dry season by higher VPD; while soil water deficit is in anti-phase with sunlight and mainly controls the second litterfall peak of the bimodal phenology in EBF. Our findings provide an important reference for modeling tropical phenology in Earth system models.
摘要
本研究收集了来自中国热带/亚热带常绿林共85个站点的凋落物量季节性变化数据,并采用线性回归、结构方程模型构建以及相位差分析等方法,综合探究中国热带/亚热带地区常绿阔叶林和针叶林叶片脱落对土壤水分、饱和水气压差和辐射强度等气候因子的响应机制。研究结果显示,在雨热同期和雨热异期两种热带/亚热带气候类型中,呈现出两种典型凋落物的物候类型(单峰季节型/双峰季节型)。在雨热同期气候条件下,光照强度和降水呈现季节性正相关,单峰的凋落物峰值和双峰的第一个峰值约出现在3–4月,不断增加的光照能促进新叶的萌发,老叶被代谢更强的新叶所替代,该类型属于一种最大程度利用光照来实现树木生长的自适应策略。双峰的第二个峰值出现在雨季末期,约在8–10月,是由不断增强的水分亏缺所导致的(常绿阔叶林:大气水分亏缺;常绿针叶林:土壤水分亏缺),这种类型是一种凋落老叶减少水分丢失来应对水分胁迫的自适应策略。在雨热异期气候条件下,光照强度和降水呈现季节性负相关,饱和水气压差与光照强度表现出一致的季节性动态变化,诱导了常绿阔叶林单峰和双峰物候的第一个凋落峰(约在3–4月),是一种权衡大气干旱胁迫和最大程度利用光照进行生长的综合自适应策略。在雨季初期,显著的土壤水分亏缺加速叶片凋落,诱导了常绿阔叶林双峰物候的第二个凋落峰(约在11月),属于凋落老叶应对土壤水分胁迫的自适应策略。这些研究结果可以为地球系统模式中热带物候的精确模拟提供重要参考。
INTRODUCTION
Litterfall is an important component of the carbon cycle in forest ecosystems and a reliable index of forest biological production and energy flow (Bray and Gorham 1964; Klemmedson et al. 1990; Maguire 1994). It is also the major pathway for the return of soil nutrients from the plant community to the soil surface in the form of organic matter, playing an important role in the maintenance of biodiversity and sustainable forest development (Domke et al. 2016; Fu et al. 2017; Vasconcelos 2004; Zhang et al. 2014a).
As a part of carbon, water and nutrient circulation processes, litterfall has received significant attention. The main emphasis in earlier litterfall studies was only placed on its amount, composition (Bray and Gorham 1964; Chandler 1941; Viro 1955; Wang 1989) and distribution (Kittredge 1944). During the past decades, scientists have made great efforts from the initial research on quantity (Bray and Gorham 1964; Finér 1996; Jia et al. 2016) to the research on quality (Brando et al. 2010; Jia et al. 2020; Zhou et al. 2006) and recently to underlying controlling factors (Fu et al. 2017; Tang and Dubayah 2019) to provide reliable phenology triggers in ecosystem models (Xiao et al. 2005).
However, the seasonal variations and driving factors of litterfall differ significantly in tropical forests, and there are still many uncertainties. In tropical wet forests, litterfall peaks were found to occur in different seasons across sites, and the temperature is unlikely to be a major limiting factor (Chen et al. 2020; De Weirdt et al. 2012; Liu et al. 2017; Tang and Dubayah 2019). These differ significantly from temperate forests. The litterfall in temperate forests in autumn reflects a cascade of senescence processes driven by ontogeny and by a shorter day length and colder temperature, which together induce a rise in abscisic acid that triggers leaf coloring and litterfall (Fracheboud et al. 2009; Keskitalo et al. 2005; Lim et al. 2007). However, tropical forests exhibit different leaf shedding strategies in response to climate and environmental seasonality. These strategies depend on soil water, atmospheric vapor pressure deficit (VPD) and shortwave incoming solar radiation (SWdown). Some studies suggested that leaf shedding is an adaptive response strategy to soil water deficit (Asner and Alencar 2010; Brando et al. 2010; Davidson et al. 2012) or atmospheric aridity (Lee and Boyce 2010; Myers et al. 1998; Xu et al. 2017; Zhang et al. 2014a). However, data from Amazon tropical forests suggested that leaf shedding before the dry seasons constitutes an adaptive strategy to replace old leaves with efficient young leaves to maximize photosynthesis when the radiation increases in the dry seasons (Chen et al. 2020, 2021; Tang and Dubayah 2019; Wright and van Schaik 1994; Wu et al. 2016; Xiao et al. 2005; Yang et al. 2021). In summary, litterfall is mainly driven by soil water availability, atmospheric water demands and sunlight availability, but it remains elusive where these strategies are dominant on a large spatial scale across the pantropic.
Here, we hypothesize that the seasonal covariance between shortwave incoming radiation and precipitation (Pre) (grouped as synchronous and asynchronous climates, respectively, see Materials and Methods) controls seasonal variation in soil water supply, atmospheric demand for water and, consequently, ecosystem photosynthesis in forests. Observations of forest litterfall in China began in the 1960s. We collected regional evergreen forest litterfall data across 85 sites in tropical and subtropical areas of China (Fig. 1) and analyzed the data separately for synchronous and asynchronous climates. The objective of the study is to comprehensively investigate under what circumstances litterfall is controlled by soil water availability, atmospheric water demands and sunlight availability. The seasonal variations in precipitation minus evapotranspiration (Pre-ET), VPD and SWdown were examined as proxies of soil water availability, atmospheric water demands and sunlight availability to analyze their relationship with the seasonal variations in forest litterfall across tropical and subtropical forests in China. The enhanced vegetation index (EVI) and near-infrared reflectance of terrestrial vegetation (NIRv) were selected as proxies of canopy photosynthesis, in addition to field litterfall data, to analyze the seasonal variations in tropical forests phenology. We expect to comprehensively understand the underlying climatic triggers of tropical and subtropical evergreen broad-leaved forests (EBFs) and evergreen needle-leaved forests (ENFs) in China and provide basic information and scientific references for the carbon cycle and nutrient cycle mechanisms of forest ecosystems in the context of global change.
MATERIALS AND METHODS
Datasets
Litterfall
Due to the gradual establishment of China’s biodiversity research network, litterfall in different types of forest ecosystems in China has been studied, ranging from rainforests in southern tropical areas to needle-leaved forests in temperate areas in northern China (Jia et al. 2016). We collected average monthly litterfall data of tropical and subtropical areas in China from published articles and screened the data from 85 studies. To ensure the data quality, we eliminated litterfall and collected data less than a year, and removed the collection period with prominent natural disturbance events or artificial control experimental data. After the screening, we obtained 85 groups of data, including 48 groups of EBFs and 37 groups of ENFs (Supplementary Table S1). The geographical positions of each data point are shown in Fig. 1.
Climate datasets
We calculated SWdown and Pre seasonality from the Climatic Research Unit–National Centers for Environmental Prediction (CRUNCEP) version 7 dataset (http://rda.ucar.edu/datasets/ds314.3/; Ryu et al. 2017) to represent the climatic light and precipitation interannual coupling relation. This dataset is a combination of CRU TS3.2 monthly data (1901–2002) and NCEP reanalysis 6-hourly data (1948–2016) with a temporal resolution of 6 hours and a spatial resolution of 0.5°. Here, the 6-hourly 0.5° SWdown, incorporating cloud cover from 1980 to 2015, was obtained from CRUNCEP version 7, provided by Laboratoire des Sciences du Climat et de l’Environnement (LSCE). The land surface ET is a central process in the climate system and a nexus of the water, energy and carbon cycles. Zhang et al. (2010) applied the normalized difference vegetation index (NDVI)-based Et algorithm with daily National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis surface meteorology and NASA/GEWEX SRB solar radiation inputs to calculate daily satellite-based ET globally. Here, we used the daily time-series ET dataset products from Zhang et al. (2010) from 1983 to 2013.
In this study, we used a 0.125° spatial resolution ERA-Interim VPD dataset, a reanalysis product based on the Integrated Forecast System of the European Centre for Medium-Range Weather Forecasts (ECMWF-IFS; Dee et al. 2011). The VPD calculation (Dee et al. 2011) follows:
where SVP and AVP are the saturated vapor pressure and actual vapor pressure (hPa), respectively. Ta and Td are the land air temperature (°C) and dew point temperature (°C), respectively.
where Z is the altitude (m). Pmst is the air pressure (hPa), and Pmsl is the air pressure at mean sea level (1013.25 hPa).
Photosynthesis proxies
Based on the geographic coordinates of the 85 litterfall observation sites, we extracted climatic factors, including Pre, Pre-ET, SWdown, VPD, EVI and NIRv, from the raster datasets for each site.
EVI was calculated from red, blue and near-infrared bands. Here, we used 0.05-degree monthly MODIS EVI products (MYD13C2, https://modis.gsfc.nasa.gov/data/dataprod/mod13.php; Badgley et al. 2017), which were built on daily atmosphere-corrected bidirectional surface reflectance (MOD09). This dataset minimized canopy-soil variations, improved sensitivity over densely vegetated regions and removed low-quality pixels using a MODIS-specific compositing method.
NIRv is a new structural parameter of the total scene near-infrared reflectance (NIR) multiplied by the NDVI. The NIRv, which was strongly correlated to the fraction of photosynthetically active radiation (fPAR), represents the proportion of pixel NIR reflectance attributable to the vegetation in the pixel. Badgley et al. (2017) evaluated that the MODIS NIRv was strongly correlated with site-level and globally gridded gross primary productivity (GPP) products. Here, we used the 0.5-degree half-monthly MODIS NIRv products provided by Badgley et al. (2017) for model output evaluation.
Method to classify the synchronous and asynchronous climates
We calculated Pearson correlation coefficients between monthly means of precipitation and SWdown (Rclimate) to quantify the differences in seasonal sunlight and water availability. Rclimate can be positive and negative, respectively:
When Rclimate >0, the region is where sunlight and moisture co-vary positively, and we refer to the region as a synchronous type. The light and moisture potentially limit forests in the same season, usually both in the drier season.
When Rclimate <0, the region is where sunlight and moisture co-vary negatively, and we refer to the region as an asynchronous type. The light and moisture potentially limit forests in different seasons, usually in the wetter and in drier season, respectively.
Method to differ unimodal and bimodal forests
We used variance-based anomaly testing to distinguish between the unimodal and bimodal phenology. We calculated the multiyear monthly mean litterfall production and corresponding variance in the monthly mean. In statistics, the variance is used to represent the difference between each value and the population mean. The greater the variance is, the greater the degree of variable dispersion. We examined the maximum variance in litterfall amount at 12 months and also calculated the slope of litterfall. If there is only one peak of variance, with a positive slope in former month and a negative slope in next month, the site is considered as a unimodal phenology. If there are two peaks of variance and the months of the two peaks happen beyond 4 months, with a positive slope in former month and a negative slope in next month, the site is considered as a bimodal phenology.
Other analysis methods
Simple linear regression
We used a simple linear regression to analyze the sunlight-rainfall coupling relationship across tropical/subtropical China and explore the relationship between the seasonality of environmental factors and seasonality of litterfall and photosynthesis proxies. All the figures and analyses were done in Python (version number 3.7).
The change rate of climatic and phenological factors
In order to further explore the impact of climate factors on litterfall, we calculated the seasonal change rates of both climatic and phenological factors using the following formula:
The rate of change was calculated and graphed in Python (version number 3.7). We took the absolute value to compare the change rates between climatic and phenological factors.
Structural equation model between climatic triggers and litterfall
Based on SPSS AMOS 21.0 software, a structural equation model (SEM) was used to analyze the magnitude and direction of environmental factors impact on the litterfall of tropical and subtropical evergreen forests in China. According to the results researched before, we quantified climate conditions using a latent variable representing VPD, Pre-ET and SWdown. These three factors can directly affect litterfall, and VPD can be influenced by Pre-ET and SWdown. Based on this logic, we set the path and SEM, and then we used AMOS to analyze the relations between climatic factors and litterfall. The resulting path coefficient showed that the level of the affected factor’s action on the amount of litterfall was deduced and calculated.
The path coefficient was used to compare the direction and size of the direct effects of different paths, and Microsoft Visio software was used to make graphs.
Phase analysis between climatic triggers and litterfall
To explore the dominant climatic variable in controlling the seasonal variation in litterfall, we applied a phase analysis of seasonal variabilities between climatic factors and litterfall described by Bradley et al. (2011) to evaluate the coincidence of seasonal cycles of VPD, SWdown, Pre and litterfall. Bradley et al. (2011) applied the cross-spectral analysis method to compare the phase differences phenology indicators with climatic factors (like sunlight and rainfall etc.). The cross-spectral analysis can be used to establish the average relative timing or phase value (‘lead’ or ‘lag’) of the climatic and phenological variables. For sites where there is an annual cycle in both the phenology time-series and a driver, the cross-spectral analysis is used to simultaneously estimate the coherency (i.e. degree of correlation at specific frequencies; range: 0–1) and timing difference (termed phase difference or often simply ‘phase’). Here, we first detected the highest values (i.e. peak) of climate variables and litterfall and then compared whether the peak of a climate variable was in-phase, leads or lags that of litterfall in the following criteria:
Variable and litterfall are ‘in phase’: the phase value is indistinguishable, with phase error equaling 0 months
Variable ‘leads’ litterfall: a positive phase that is between 1 and 5 months
Variable ‘lags’ litterfall: a negative phase that is between −1 and −5 months
Variables with “in-phase” and “1–5 months” litterfall were detected as drivers of litterfall.
RESULTS
Forest litterfall discrepancies in tropical/subtropical China
The 48 groups of litterfall in the EBFs and 37 groups in ENFs are listed in Supplementary Table S1, distributed in Fig. 1a and the production of all sites are shown in Supplementary Fig. S1. The results show that the mean annual production of litterfall of broad-leaved forest is 6566 ± 2684 kg hm−2, ranging from 1712 to 14 180 kg hm−2, mainly concentrated in the range of 6000–8000 kg hm−2 (39.58%, Fig. 1b). The mean annual production of litterfall of needle-leaved forest is 4410 ± 2300 kg hm−2, ranging from 531 to 10 523 kg·hm−2, and mainly concentrated in the range of 2000–6000 kg hm−2 (40.54%, Fig. 1b). In general, the average annual litterfall production in EBF is significantly higher than that in ENF.
According to covariances between seasonal sunlight and water availability, 71 sites are in a synchronous climate, where the correlations between seasonal sunlight and rainfall are positive (Rclimate > 0), of which there are 37 EBF types and 34 ENF types. In addition, 14 sites are in an asynchronous climate, where the correlations between seasonal sunlight and rainfall are negative (Rclimate < 0), of which there are 11 EBF types and 3 ENF types. According to the number of litterfall peaks in the seasonality curve, the litterfall phenology is grouped into unimodal and bimodal types (Materials and Methods). In general, 36 of 85 tropical forests show a unimodal type (23 EBFs and 13 evergreen coniferous forests), while 41 of 85 tropical forests show a bimodal type (22 EBFs and 19 ENFs). The remaining eight sites show no regular seasonal cycle. Based on this, we comprehensively investigated the seasonal controls of precipitation, Pre-ET, VPD and SWdown on the unimodal and bimodal phenology of forest litterfall in EBF/ENF types under synchronous/asynchronous climates.
Synchronous climate (Rclimate > 0)
Evergreen broad-leaved forests
In total, 56.4% of the total sites are located in a synchronous climate. We found that litterfall in most EBFs shows two types of seasonality: unimodal litterfall and bimodal litterfall. Figure 2a shows that the peak of the unimodal type occurs during April, at the end of the dry season with low precipitation and high radiation. The SEM model analysis suggests that for the unimodal type of EBF, SWdown and Pre-ET have significantly positive effects on litterfall seasonality (r = 0.34, P < 0.05; r = 0.45, P < 0.05), while VPD has a slightly significant effect on litterfall (Fig. 5a). Phase analysis indicates that SWdown, VPD and precipitation are in phase with litterfall seasonality for 76%, 42% and 58% of the sites, respectively (Fig. 4a). This peak value is caused by gradually increased sunlight input. The change rate is higher than the average value in the peak month (Fig. 3a). We named this type rejuvenation one, suggesting a light-controlled phenology strategy to maximize radiation use by dropping old leaves and simultaneously making young leaves with higher photosynthetic efficiency.
Figure 2b presents the bimodal type of EBF. The two peaks occur separately during April and August. The phase analysis on the annual seasonality shows that SWdown, VPD and precipitation are in phase with litterfall seasonality for 64%, 58% and 35% of the sites, respectively (Fig. 4b). For first litterfall in April, the climate factors (SWdown, Pre and VPD) and vegetation proxies (EVI and NIRv) all increase rapidly (Fig. 2a). The SEM model analysis shows that SWdown significantly affected litterfall (r = 0.8, P < 0.05), but VPD had little effect on litterfall (r = 0.21, P < 0.05, Fig. 5b). A higher rate of sunlight increase can affect tree sprouts (Fig. 3b, Zhang et al. 2014b). We assumed that the first peak of litterfall is caused by replacing old leaves with young leaves for canopy rejuvenation. For the second peak value of litterfall in August, Fig. 5c shows that VPD (r = 0.3, P < 0.05) and Pre-ET (r = 0.08, P < 0.05) had significantly positive effects on litterfall. Therefore, we speculated that the second peak of litterfall is caused by atmospheric water stress due to high values of VPD, which can increase abscisic acid and lead to an increase in litterfall and the effect of typhoon. Additionally, typhoons are concentrated during this time period, which can also cause a large amount of falling leaves, branches and fruits, and a peak in litterfall (Sato 2004).
Evergreen needle-leaved forests
In total, 47.8% of ENF in the synchronous climate shows a unimodal type in litterfall seasonality (Fig. 2c). Unlike the unimodal type of EBF, whose peak occur at the end of the dry season, most ENFs peaks in leaf shedding during the rainy season, when heat and rain conditions are optimal. Phase analysis (Fig. 4c) shows that the seasonality of VPD and SWdown are both in phase with litterfall in 66% of sites, but precipitation lags one month with litterfall. At the same time, additional SEM model analysis (Fig. 5d) shows that SWdown, VPD and Pre-ET all show great potential in triggering off litterfall seasonality (r = 0.29, 0.36, 0.20; P < 0.05, respectively).
Figure 2d shows ENF with a bimodal monthly pattern of litterfall. The first peak is similar to that of EBF. The high rate of SWdown likely induces a replacement of old leaves by new leaves for canopy rejuvenation (Fig. 3d) and triggers off a large amount of litterfall in April. Although Fig. 5e shows that Pre-ET might also negatively affect litterfall (r = −0.19, P < 0.05), litterfall increases when soil moisture decreases, suggesting a noncausal correlation between litterfall and soil water stress. SWdown is a more reasonable trigger for the first peak (r = 0.56, P < 0.05). However, for the second peak of litterfall, the SEM model (Fig. 5f) indicates a negative relationship between SWdown and litterfall (r = −0.35, P < 0.05) and a negative relationship between VPD and litterfall (r = −0.26, P < 0.05). And Pre-ET thereby is a more reasonable trigger for the first peak on litterfall (r = −0.29, P < 0.05), when there is a stronger soil water stress as the Pre-ET becomes lower. At the month around November, the precipitation has the lowest value of 47 mm mol−1 (mean = 144 mm mol−1), which led to a soil water deficit that promoted abscisic acid synthesis and triggered litterfall. Thus, the second peak is mainly due to the soil water deficit.
Asynchronous climate (Rclimate< 0)
Evergreen broad-leaved forests
In contrast to the synchronous climate where VPD, soil moisture and SWdown peak in the same month, the asynchronous climate encounters soil water deficit, atmospheric water deficit and sunlight inhabitation in different months. The values of VPD and SWdown are both relatively higher (1146 ± 53 Pa, 246 ± 1 W m−2) in months around March (Fig. 2e), and the VPD and SWdown seasonality are in phase with that of litterfall at 62.5% and 87.5% of sites, respectively (Fig. 4e). The SEM analysis generally shows that SWdown and VPD significantly affected litterfall (r = 0.35, r = 0.34, P < 0.05, Fig. 5g). The climate factors (SWdown, precipitation and VPD) and vegetation proxies (EVI and NIRv) all increase rapidly (Fig. 2e). The change rate of precipitation is lower in March (Fig. 3e). After that, the rate increased rapidly and then caused a declining trend for litterfall production. Therefore, we attribute the litterfall peak as the stress type due to high atmospheric water demands caused by a higher VPD.
For the bimodal type (Fig. 2f), both SWdown and VPD act as potential litterfall triggers, as the phases of SWdown and VPD seasonality are consistent (Fig. 4f). During the period of the first litterfall peak, the SEM model also shows that SWdown and VPD have path coefficients of litterfall of 0.61 and 0.66 (P < 0.05), respectively (Fig. 5h). However, the high increasing rate of SWdown lags behind litterfall seasonality, suggesting that SWdown is not a reasonable trigger for the first peak of litterfall. We attribute the litterfall peak to water stress from a high VPD (Fig. 2e). In contrast, the second peak around November is attributed to the soil water deficit. During this period, the precipitation is at the lowest value (83 mm mol−1), and the path coefficient between Pre-ET and litterfall is −0.31 (P < 0.05, Fig. 5i), which can lead to low soil moisture and directly boost the rate of litterfall.
Evergreen needle-leaved forests
The unimodal ENF in the asynchronous type has a high peak value, reaching 2011.2 kg hm−2 in approximately February (Fig. 2g), when Pre is 28.5 mm mol−1 (mean = 149 mm mol−1), which is much lower than the annual average. The SEM results show that the three environmental factors have no significant effect (P > 0.05) on litterfall production (Fig. 5j). However, VPD and SWdown led to litterfall by 2 months and 1 month, respectively. Similarly, for the bimodal type, the SEM model also shows that the three environmental factors are not significantly related to litterfall production (Fig. 5k and l). These results show that the litterfall pattern might be influenced jointly by three climatic factors rather than by a strong single factor.
DISCUSSION
Vegetation productivity and leaf longevity influence the discrepancies in litterfall production
Litterfall production in different types of forest biome fluctuate considerably (Bray and Gorham 1964; Kavvadias et al. 2001; Liu et al. 1997; Pedersen and Hansen 1999; Sundarapandian and Swamy 1999; Zhou et al. 2006). Vegetation productivity is seen as one of the most important factors and shows considerable impacts on litterfall production (Reich et al. 1992). For examples, the broad-leaved trees have more branches with thick trunks and large crown widths (Ning et al. 2009), and consequently generate higher litterfall production than needle-leaved trees (Dan and Michael 1998; Zhou et al. 2006). We also found that the litterfall productions of EBFs, with larger values of GPP and vegetation indices, including leaf area index (LAI), NIRv, NDVI and EVI (Table 1), are on average larger than those of needle-leaved forests (Fig. 1b). The analyses in this study support above hypothesis of vegetation productivity impacts on the discrepancies in litterfall production.
. | Litterfall (kg·hm2) . | LAI . | NIRv . | NDVI . | EVI . | GPP . |
---|---|---|---|---|---|---|
EBF | 6566 ± 2684 | 2.64 | 0.16 | 0.65 | 0.38 | 5.5 |
ENF | 4410 ± 2300 | 1.87 | 0.14 | 0.59 | 0.33 | 4.8 |
. | Litterfall (kg·hm2) . | LAI . | NIRv . | NDVI . | EVI . | GPP . |
---|---|---|---|---|---|---|
EBF | 6566 ± 2684 | 2.64 | 0.16 | 0.65 | 0.38 | 5.5 |
ENF | 4410 ± 2300 | 1.87 | 0.14 | 0.59 | 0.33 | 4.8 |
. | Litterfall (kg·hm2) . | LAI . | NIRv . | NDVI . | EVI . | GPP . |
---|---|---|---|---|---|---|
EBF | 6566 ± 2684 | 2.64 | 0.16 | 0.65 | 0.38 | 5.5 |
ENF | 4410 ± 2300 | 1.87 | 0.14 | 0.59 | 0.33 | 4.8 |
. | Litterfall (kg·hm2) . | LAI . | NIRv . | NDVI . | EVI . | GPP . |
---|---|---|---|---|---|---|
EBF | 6566 ± 2684 | 2.64 | 0.16 | 0.65 | 0.38 | 5.5 |
ENF | 4410 ± 2300 | 1.87 | 0.14 | 0.59 | 0.33 | 4.8 |
In addition to vegetation productivity, leaf longevity is another important factor influencing amounts of litterfall production (Kamruzzaman et al. 2016). Many studies have observed different leaf longevities among different tree species (Wang et al. 2000; Zhang 2004; Zhang et al. 2016). For examples, the EBFs usually have a leaf life span of only 1–2 years (Reich et al. 1992; Wang et al. 2000). While the needle leaves of evergreen forests can remain on the branch for 2–3 years and some can even survive for 4–7 years (Wang et al. 2000; Zhou et al. 2008). Thus, in the evergreen biome, needle-leaved forests tend to have longer leaf life spans than broad-leaved plants (Zhou et al. 2008), both of which are usually longer-lived than that of deciduous plants (Williams et al. 1989). Herein, we collected the average leaf longevity of evergreen broad-leaved and needle-leaved trees from former literatures in tropical and subtropical China (Supplementary Table S2; Zhou et al. 2008). Our results support that longer leaf life in ENFs likely leads to a smaller litterfall production than evergreen broad-leaved plants with a shorter leaf life span.
Sunlight and water stress separately dominate dry- and wet-season litterfall under synchronous climate
Former studies concluded two main types of leaf shedding strategies of tropical and subtropical forests. One is an adaptive response strategy to minimize costs under water stress and to void hydraulic failure in dry seasons (Asner and Alencar 2010; Brando et al. 2010; Davidson et al. 2012; Lee and Boyce 2010; Myers et al. 1998; Xu et al. 2017; Zhang et al. 2014b). One is an adaptive strategy to replace old leaves with efficient young leaves to maximize photosynthesis when the radiation increases in the dry seasons (Chen et al. 2020; Tang and Dubayah 2019; Wright and van Schaik 1994; Wu et al. 2016; Xiao et al. 2005). However, it is still unclear in which climate, soil water deficit, atmospheric water deficit and sunlight inhibition act separately or synergistically to control litterfall seasonality.
Former studies mainly explored the constraints of litterfall seasonality at annual cycle. However, the seasonal dynamics of forest litterfall in large parts of tropical and subtropical China show distinctly bimodal pattern (Liu et al. 2017; Ning et al. 2009). The first litterfall peak always happened at the end of the dry season from March to April ( Pan et al. 2010; Ren et al. 1999; Tang 2010). The second litterfall peak mainly occurred at the end of the rainy season from August to October (Guo et al. 2006; Zhu et al. 2021). Former studies overlooked this points that the underlying climatic triggers of litterfall in these two periods might differ greatly. Our analysis confirmed that the, for litterfall peak at the end of dry season, the increasing sunlight input mainly contributed to the litterfall rhythm of evergreen trees, causing new leaf flushing to replace old leaves (Fig. 6a). This type of leaf shedding is an adaptive strategy to replace old leaves with efficient young leaves (Chen et al. 2020; Tang and Dubayah 2019; Wright and van Schaik 1994; Wu et al. 2016; Xiao et al. 2005). Plants aim to maximize canopy photosynthesis in response to increasing radiation because young leaves have a higher light use efficiency (Carswell et al. 2002; De Weirdt et al. 2012) or to increase the survival rate of young leaves in the dry season when herbivory and fungal pressure are lower (Wu et al. 2016). For the second litterfall peak at the end of the rainy season, our analyses show that the atmospheric water deficit controls the litterfall peak of EBF, while the soil moisture deficit controls that of ENF (Fig. 6b). This type of leaf shedding is an adaptive strategy to prevent water stress damage to plant hydraulic systems from soil water deficit (Asner and Alencar 2010; Brando et al. 2010; Davidson et al. 2012; Liu et al. 2021) or atmospheric water deficit (Lee and Boyce 2010; Myers et al. 1998) and to avoid high maintenance respiration costs and reduce transpiration at the cost of reduced photosynthesis (Xu et al. 2017; Zhang et al. 2010). Soil moisture deficits impede the plant hydraulics’ movement of water from the soil to the leaf (Tyree and Sperry 1989) and result in soil water deficit (Novick et al. 2016). However, even with adequate soil water to transpire, tropical trees might encounter another independent stress to trigger litterfall from increased atmospheric demand, i.e. higher VPD, the difference between saturated and actual water vapor (Konings et al. 2017; Novick et al. 2016).
Water stress is the main constraint of litterfall under an asynchronous climate
Under the asynchronous climate (Rclimate < 0), the seasonal constraints of water and sunlight happen during the same period (i.e. dry seasons) across tropical and subtropical China. The litterfall peak of the unimodal type and the first peak of the bimodal phenology are always at the end of the dry season from March to April (Zhu et al. 2021), similar to the synchronous climate. Increasing sunlight radiation results in higher temperature and longer sunshine hours, which can promote tree sprouting, with more young leaves having a higher light use efficiency (Carswell et al. 2002; De Weirdt et al. 2012; Wright and van Schaik 1994). Meanwhile, the VPD increases as a result of increasing temperature, which show consistent seasonal variation with sunlight. This can enhance atmospheric dryness and abscisic acid levels in leaves to accelerate leaf senescence and shedding (Lee and Boyce 2010; Reichstein et al. 2002) (Fig. 6c). Seasonal precipitation is one of the most important factors and less precipitation will cause soil water deficit to accelerate leaf shedding and, to a large extent, determines the deciduous and leaf-spreading periods of plants (Jorgensen et al. 1975; Li et al. 2021; Linkosalo et al. 2006; Wang 2010; Zhu et al. 2021). The second litterfall peak of the bimodal phenology is due to soil water deficit caused mainly by decreases in precipitation, following an adaptive strategy to prevent water stress damage to plant hydraulic systems from soil water deficit (Asner and Alencar 2010; Brando et al. 2010; Davidson et al. 2012).
CONCLUSIONS
In summary, this study provides general insight into characterizing the phenology adaptations of tropical and subtropical forests to climate seasonality regimes in tropical and subtropical China. We found that forest biomass and leaf longevity control the quantity differences between EBFs and ENFs. The seasonal dynamics of litterfall are relatively complex. We concluded that two main patterns of seasonal litterfall (unimodal and bimodal ones) of exist under two types of climates (sunlight and rainfall synchronous/asynchronous climates). The analyses from 71 sites show that sunlight triggers off the litterfall peak at the end of the dry season and that water stress leads to a litterfall peak at the end of the wet seasons under a synchronous climate, where seasonal sunlight and rainfall are positively correlated. Under an asynchronous climate, where seasonal sunlight and rainfall are negatively correlated, water stress mainly puts constraints on leaf shedding for EBF. However, the driving mechanisms are more complex for ENF under asynchronous climate. Notably, under some specific conditions, the seasonal litterfall of evergreen trees might be affected by other extreme climatic events (Deng et al. 2017; Wang et al. 2014), which lead to multiple seasonal litterfall peaks. For example, many studies have observed litterfall peaks during typhoon events or extreme rainfall (Sato 2004).
In conclusion, this study advances the understanding of the environmental triggers of tropical and subtropical evergreen forest leaf phenology and provides an important reference for modeling tropical and subtropical phenology in Earth system models. The background climate and specific meteorological disturbances can affect the rhythm of forest litterfall, resulting in a diversity of seasonal dynamics at different scales. More well-designed manipulation experiments in a large spatial region are needed in the future to test the proposed adaptive strategies of tropical and subtropical forest leaf shedding to climate.
Supplementary Material
Supplementary material is available at Journal of Plant Ecology online.
Table S1: The information of all data.
Table S2: Life expectancy (EX) of main species.
Figure S1: Histogram of litterfall production for all 85 data points.
Funding
This study was supported by the National Natural Science Foundation of China (grant numbers 31971458 and 41971275), Special High-Level Plan Project of Guangdong Province (grant number 2016TQ03Z354), Forestry Science and Technology Innovation Project of Guangdong Province (grant number 2021KJCX003) and ‘GDAS’ Project of Science and Technology Development (grant numbers 2020GDASYL-20200302001 and 2020GDASYL-20200102002).
Conflict of interest statement. The authors declare that they have no conflict of interest.